《電子技術(shù)應(yīng)用》
您所在的位置:首頁 > 其他 > 設(shè)計(jì)應(yīng)用 > 卷積神經(jīng)網(wǎng)絡(luò)識別地基云圖的數(shù)據(jù)庫建立及處理方法
卷積神經(jīng)網(wǎng)絡(luò)識別地基云圖的數(shù)據(jù)庫建立及處理方法
2020年信息技術(shù)與網(wǎng)絡(luò)安全第3期
王敏1,2,,周樹道1,2,,劉展華1,,任尚書3
(1.國防科技大學(xué) 氣象海洋學(xué)院,,江蘇 南京 211101; 2.南京信息工程大學(xué) 氣象災(zāi)害預(yù)警與評估協(xié)同創(chuàng)新中心,,江蘇 南京 210044,; 3.解放軍95171部隊(duì),廣東 廣州 510000)
摘要: 卷積神經(jīng)網(wǎng)絡(luò)(Convolutional neural network,CNN)具有非比尋常的從樣本中學(xué)習(xí)特征的能力,,訓(xùn)練需要大量帶有標(biāo)簽的圖像樣本,。因此,在使用卷積神經(jīng)網(wǎng)絡(luò)對地基云圖相關(guān)研究時(shí),,建立云圖樣本庫是第一步,,也是非常重要的一步。首先,,通過數(shù)碼相機(jī)直接拍攝,、從互聯(lián)網(wǎng)上下載、從公開發(fā)行的云圖類書籍獲取以及由全天空照相機(jī)拍攝等手段獲取三個(gè)云圖樣本庫,;接著,,對三個(gè)樣本庫圖像的分辨率、噪聲,、數(shù)量等問題進(jìn)行了分析,;然后,,采用雙線性插值和數(shù)據(jù)增強(qiáng)方法對樣本庫進(jìn)行歸一化預(yù)處理,;最后,利用卷積神經(jīng)網(wǎng)絡(luò),、LBP,、Heinle feature和Textonbased method三種方法對增強(qiáng)后的數(shù)據(jù)集進(jìn)行云識別分類驗(yàn)證,實(shí)驗(yàn)結(jié)果表明,,利用本文方法進(jìn)行增強(qiáng)數(shù)據(jù)可有效解決卷積神經(jīng)網(wǎng)絡(luò)對小樣本數(shù)據(jù)識別率不高
中圖分類號:TP412.15
文獻(xiàn)標(biāo)識碼:A
DOI: 10.19358/j.issn.2096-5133.2020.03.011
引用格式:王敏,,周樹道,劉展華,,等.卷積神經(jīng)網(wǎng)絡(luò)識別地基云圖的數(shù)據(jù)庫建立及處理方法[J].信息技術(shù)與網(wǎng)絡(luò)安全,,2020,39(3):56-61.
Establishment and processing of Groundbased cloud image database for CNN
Wang Min1,2,Zhou Shudao1,2,Liu Zhanhua1,Ren Shangshu3
(1.College of Meteorology and Oceanography,National University of Defense Technology,Nanjing 211101,China; 2.Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science & Technology,Nanjing 210044,China; 3.Unit 95171 of PLA,Guangzhou 510000, China)
Abstract: Convolutional neural network (CNN) has an extraordinary ability to learn features from samples,and training requires a large number of image samples with labels.Therefore,it is the first and very important step to establish cloud image sample bank when using convolutional neural network to study the groundbased cloud image.Firstly,three cloud image sample libraries are acquired by means of digital camera direct shooting,downloading from the Internet,acquiring from publicly released cloud image books,and shooting by allsky camera.Then,the resolution,noise and number of images in the three sample libraries are analyzed.Then,bilinear interpolation and data enhancement are used to normalize the sample database.Finally,CNN,LBP,Heinle feature and Textonbased method are used to verify the cloud recognition of the enhanced data set.The experimental results show that the improved data can effectively solve the problems of convolution neural network for small sample data recognition such as low rate and incomplete network operation, and lays a foundation for the application of convolutional neural network in the recognition of groundbased cloud image.
Key words : convolutional neural network;supervised learning;sample bank;normalization

0    引言

云是地球上水文循環(huán)的一個(gè)重要環(huán)節(jié),它與地面輻射相互作用共同影響著局地和全球尺度的能量平衡,。云分類對天氣預(yù)報(bào)很重要,,直接決定著降水,、降雪、雹和雷電等天氣活動,。地基云觀測數(shù)據(jù)主要包括云量,、云狀、云底高度,,根據(jù)三者的不同表現(xiàn)可以將云分為3族,、10屬、29類,,具有種類多,、變化快、相似,、易與天空背景融合等特點(diǎn),。實(shí)際觀測中人工觀測為主,存在著主觀性強(qiáng),、準(zhǔn)靜態(tài),、成本高、觀測點(diǎn)偏少以及信息記錄不完整等問題,。目前的地基云圖云狀自動化識別方法通常采用圖像預(yù)處理→特征提取→分類器分類這樣的流程,。

大多數(shù)研究者重點(diǎn)研究表達(dá)不同云屬性的特征提取技術(shù),但這種識別分類方法是基于人工經(jīng)驗(yàn)提取特征的,,且各個(gè)環(huán)節(jié)都是獨(dú)立的,,只有簡單的兩三層學(xué)習(xí)網(wǎng)絡(luò),實(shí)則是一種“淺層學(xué)習(xí)”,,致使此類方法適用的云類別范圍有限,,加之分類器的選取、云的復(fù)雜變化,,影響了器測云狀識別的識別精度及識別速度,,僅能簡單識別積云、層云,、高積云,、卷云等少數(shù)四至五類典型云的自動識別。




本文詳細(xì)內(nèi)容請下載:http://forexkbc.com/resource/share/2000003184





作者信息:

王敏1,2,,周樹道1,2,,劉展華1,任尚書3

(1.國防科技大學(xué) 氣象海洋學(xué)院,,江蘇 南京 211101,;2.南京信息工程大學(xué) 氣象災(zāi)害預(yù)警與評估協(xié)同創(chuàng)新中心,江蘇 南京 210044;3.解放軍95171部隊(duì),,廣東 廣州 510000)


此內(nèi)容為AET網(wǎng)站原創(chuàng),,未經(jīng)授權(quán)禁止轉(zhuǎn)載。